Visualize: Layers, Exploratory Data Analysis, Communication

Module 04

Ray J. Hoobler

Layers (ggplot2)

Asthetic mapping

What can be included in the arguments of the aes() function?

ggplot(data = <DATA>) + 
  <GEOM_FUNCTION>(
     mapping = aes(<MAPPINGS>)
  )

Variables (quantitative or categorical) can be mapped to the following aesthetics:

  • x
  • y
  • color
  • fill
  • size
  • shape
  • linetype
  • alpha

Explore Aesthetics with the Palmer Penguins Data (1/6)

Code 
library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm))

Explore Aesthetics with the Palmer Penguins Data (2/6)

Code 
ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm, 
                           color = species))

Explore Aesthetics with the Palmer Penguins Data (3/6)

Code 
ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm, 
                           color = species, 
                           size = body_mass_g))

Explore Aesthetics with the Palmer Penguins Data (4/6)

Code 
ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm, 
                           color = species, 
                           size = body_mass_g, 
                           shape=island))

Explore Aesthetics with the Palmer Penguins Data (5/6)

Code 
ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm, 
                           size = body_mass_g, 
                           shape=island, 
                           alpha = 0.5),
             color = "blue")

Explore Aesthetics with the Palmer Penguins Data (6/6)

What’s happening here?

Code 
ggplot(data = penguins) + 
  geom_point(mapping = aes(x = bill_length_mm, y = bill_depth_mm, 
                           color = "blue"))

Geometric Objects (1/2)

ggplot2 provides
more than 40 geoms!

Code 
library(patchwork)

p1 <- ggplot(penguins, aes(x = body_mass_g)) + 
  geom_histogram()

p2 <- ggplot(penguins, aes(x =body_mass_g)) +
  geom_density()  

p3 <- ggplot(penguins, aes(x = body_mass_g)) +
  geom_boxplot()

p1 + p2 + p3

Geometric Objects (2/2)

Libraries are availabe for specialty plots (i.e., “pair plots”)

Code 
library(GGally)

ggpairs(penguins, columns = 3:6, (aes(color = species, alpha = 1/4)))

Facets

The pair plot above is an example of a facet plot. We can create a facet plot by adding the facet_wrap() or facet_grid() function to the ggplot object.

Code 
penguins |> 
  ggplot() +
  geom_histogram(aes(x = body_mass_g/1000, fill = species), bins = 30) +
  facet_wrap(vars(island), ) +
  theme(panel.spacing.x = unit(1, "lines"))

facet_grid() (1/3)

Code 
penguins |> 
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = flipper_length_mm, color=species)) +
  facet_grid(vars(sex), vars(island)) +
  theme(panel.spacing.x = unit(1, "lines"))

facet_grid() (1/3)

Code 
penguins |> 
  filter(!is.na(sex)) |>
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = flipper_length_mm, color=species)) +
  facet_grid(vars(sex), vars(island)) +
  theme(panel.spacing.x = unit(1, "lines"))

facet_grid() (3/3)

Code 
penguins |> 
  filter(!is.na(sex)) |>
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = flipper_length_mm, color=island)) +
  facet_grid(vars(species), vars(sex)) +
  theme(panel.spacing.x = unit(1, "lines"))

Facets using the ~ operator

Using the ~ operator to specify the rows and columns of the facet grid.

Code 
penguins |> 
  filter(!is.na(sex)) |>
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = flipper_length_mm, color=island)) +
  facet_grid(species ~ .) +
  theme(panel.spacing.x = unit(1, "lines"))

Using the Dot(.) in ggplot2 Faceting

Prompt for Claude

Can you explain the use of “.” in R? Specifically, as it’s used in ggplot2 for functions like facet_grid(. ~ species)

Response from Claude

The dot (.) in R, especially as used in ggplot2 functions like facet_grid(), has a special meaning. Let me explain its use:

  1. In ggplot2’s facet_grid():

    In the context of facet_grid(. ~ species), the dot serves as a placeholder. It indicates that you don’t want to split the plot on the rows (left side of the tilde ~), but you do want to split it on the columns (right side of the tilde) based on the “species” variable.

    • facet_grid(. ~ species): Creates separate plots for each unique value in the “species” column, arranged horizontally.
    • facet_grid(species ~ .): Would create separate plots for each unique value in the “species” column, arranged vertically.
    • facet_grid(var1 ~ var2): Would create a grid of plots, with “var1” determining the rows and “var2” the columns.

  2. In other R contexts:

    The dot can also be used in other situations in R:

    • In pipe operations (%>% or |>), the dot represents the object being passed through the pipe.
    • In formulas, it can represent “all other variables” not explicitly mentioned.
    • In function definitions, it can be used to pass additional arguments.

Assignment

Chapter 9 9.4.1 Exercises. Questions 1 through 7

Chapter 9 9.3.1 Exercises, questions 1 through 4

Statistical Transformations (1/2)

Several geometric objects in ggplot2 carry out statistical transformations, examples include:

  • geom_histogram() calculates the count of observations in bins.
  • geom_density() calculates the density of a variable.
  • geom_bar() calculates the count of observations for each level of a categorical variable.
  • geom_boxplot() calculates the median, quartiles, and outliers of a variable.
  • geom_smooth() calculates a smoothed line (or fit) through the data.

Figure 9.2 from R4DS 2e provides the following graphic:

Statistical Transformations (2/2)

For every geometric object geom_*, there is a corresponding statistical transformation stat_*.

Additional transformations can be called to override the default values. For instance, the geom_histogram() includes the following:

Computed variables

These are calculated by the ‘stat’ part of layers and can be accessed with delayed evaluation.

  • after_stat(count)
    number of points in bin.

  • after_stat(density)
    density of points in bin, scaled to integrate to 1.

  • after_stat(ncount)
    count, scaled to a maximum of 1.

  • after_stat(ndensity)
    density, scaled to a maximum of 1.

  • after_stat(width)
    widths of bins.

Code 
penguins |> 
  ggplot() +
  geom_histogram(aes(x = body_mass_g, y = after_stat(density), fill = species), 
                 bins = 30, show.legend = FALSE) +
  theme(
    axis.title.y = element_text(size = 24, color = "red", face = "bold")
  )

Position Adjustments (1/6)

Code 
machine <- read_table("datasets/MACHINE.DAT", 
                      skip = 25, 
                      col_names = c("machine", "day", "time", "sample", "diameter"), 
                      col_types = "ffffd")

machine

Position Adjustments (2/6)

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = day, y = diameter, 
                 color = machine))

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = day, y = diameter, 
                 color = machine), position = "jitter")

Position Adjustments (3/6)

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = day, y = diameter, color = machine), 
             position = position_jitter(width = 0.1))

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = time, y = diameter, color = machine), 
             position = position_jitter(width = 0.1))

Position Adjustments (4/6)

Code 
machine |> 
  mutate(machine_time = str_c(machine, time, sep = "_")) |> 
  ggplot() +
  geom_point(aes(x = machine_time, y = diameter, color=machine), position = position_jitter(width = 0.1))

Position Adjustments (5/6)

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = machine, y = diameter, color = time), position = position_dodge(width = 0.5))

Position Adjustments (6/6)

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = machine, y = diameter, color = time), position = position_jitterdodge(dodge.width = 0.5, jitter.width = 0.5))

The Layered Grammar of Graphics

```{r}
ggplot(data = <DATA>) + 
  <GEOM_FUNCTION>(
     mapping = aes(<MAPPINGS>),
     stat = <STAT>, 
     position = <POSITION>
  ) +
  <COORDINATE_FUNCTION> +
  <FACET_FUNCTION>
```

Exploratory Data Analysis (EDA)

EDA with the MACHINE.DAT Data (1/3)

Location AND Variation

Code 
machine |> 
  group_by(machine, time) |>
  summarise(mean_diameter = mean(diameter), 
            sd_diameter = sd(diameter), 
            ci_diameter = 1.96 * sd(diameter) / sqrt(n())) |>
  ungroup() |>
  mutate(machine_time = str_c(machine, time, sep = "_")) |>
  ggplot() +
  geom_point(aes(x = machine_time, y = mean_diameter, color = machine), size = 4) +
  geom_errorbar(aes(x = machine_time, 
                    ymin = mean_diameter - ci_diameter, 
                    ymax = mean_diameter + ci_diameter), width = 0.25)

EDA with the MACHINE.DAT data (2/3)

Location AND Variation

Code 
machine |> 
  ggplot() +
  geom_density(aes(x = diameter, fill = machine), alpha=1/3) +
  geom_vline(xintercept = 0.125, linetype = "dashed") +
  scale_x_continuous(limits = c(0.1175, 0.1325)) +
  facet_grid(rows = vars(time))

EDA with the MACHINE.DAT data (3/3)

Location AND Variation

Code 
machine |> 
  ggplot() +
  geom_point(aes(x = machine, y = diameter, color = time), width = 0.1, position = position_jitterdodge(dodge.width = .75, jitter.width = 0.2)) +
  geom_boxplot(aes(x = machine, y = diameter, fill = time), alpha = 1/3)

Assignment

  • Carry out an EDA on the FURNACE.DAT dataset.
  • Are there any trends in the data that would be difficult to detect with standard summary statistics?
  • Are there any outliers in the data? (A description of the experiment is available on the NIST website)
Code 
column_names <- c("RUN ZONE WAFER THICKNESS") |> 
  str_to_lower() |> 
  str_split(" ") |> 
  unlist()

furnace <- read_table("datasets/FURNACE.DAT", 
                      skip = 25, 
                      col_names = column_names, 
                      col_types = "iffd")

furnace |> 
  ggplot() +
  geom_point(aes(x = run, y = thickness))

Communication

Labels

For most plots, the labs() layer will provide the right amount of customization.

Code 
penguins |> 
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = bill_depth_mm, color = species)) +
  labs(title = "Bill Length and Depth Serve as Natural Classifiers", 
       subtitle = "palmerpenguins: Palmer Archipelago (Antarctica) Penguin Data",
       x = "Bill Length (mm)", 
       y = "Bill Depth (mm)", 
       color = "Species",
       caption = "https://github.com/allisonhorst/palmerpenguins")

Annotations

  • geom_text()

  • annotate()

Annotations: geom_text() (1/2)

Calculate mean values for Bill Depth and Bill Length for each species

Code 
mean_bill_depth_length <- penguins |> 
  group_by(species) |> 
  summarise(mean_bill_length = mean(bill_length_mm, na.rm = TRUE), 
            mean_bill_depth = mean(bill_depth_mm, na.rm = TRUE)) |> 
  ungroup()

mean_bill_depth_length

Annotations: geom_text() (2/2)

Code 
library(ggrepel)

penguins |> 
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = bill_depth_mm, color = species), alpha = 1/4, show.legend = FALSE) +
  geom_point(data = mean_bill_depth_length, 
             aes(x = mean_bill_length, y = mean_bill_depth, color = species),
             size = 4, show.legend = FALSE) +
  geom_label_repel(data = mean_bill_depth_length, 
            aes(x = mean_bill_length, y = mean_bill_depth, label = species, color = species), 
            size = 4, fontface = "bold", box.padding = 1, show.legend = FALSE, seed = 42) +
  labs(
    title = "Comparison of Means",
    subtitle = "Bill Length and Depth by Species",
    x = "Bill Length (mm)",
    y = "Bill Depth (mm)",
    caption = "palmerpenguins: Palmer Archipelago (Antarctica) Penguin Data"
  ) +
  theme_classic()

Annotations: annotate()

Code 
furnace |> 
  ggplot() +
  geom_point(aes(x = seq_along(thickness), y = thickness)) +
  geom_hline(yintercept = 560, linetype = "dashed") +
  geom_hline(yintercept = c(460, 660), linetype = "dashed", color = "red") +
  annotate("text", x = 175, y = 560-50, label = "Target Thickness", color = "black", size = 4, fontface = "bold") +
  annotate("segment", x = 175, xend = 190, y = 520, yend = 560, 
           arrow = arrow(ends = "last", length = unit(0.3, "cm")), 
           size = 0.5, color = "black") +
  annotate("text", x = 20, y = 460+10, label = "Lower Control Limit", color = "red", size = 4, fontface = "bold") +
  annotate("text", x = 20, y = 660-10, label = "Upper Control Limit", color = "red", size = 4, fontface = "bold") +
  labs(
    title = "Run Sequence Plot of Furnace Data",
    subtitle = "Thickness of processed wafers",
    x = "Sample Number",
    y = "Thickness (Å)",
    caption = "\nhttps://www.itl.nist.gov/div898/handbook/ppc/section5/ppc511.htm"
  ) +
  theme_classic()

Scales (1/4)

FRED Economic Data

Source: U.S. Census Bureau

Release: Income and Poverty in the United States

Units: 2022 CPI-U-RS Adjusted Dollars, Not Seasonally Adjusted

Frequency: Annual

Household data are collected as of March.

. . . the Consumer Price Index retroactive series using current methods (R-CPI-U-RS) presents an estimate of the CPI for all Urban Consumers (CPI-U) from 1978 to the present that incorporates, when possible, most of the improvements made over that time span into the entire series.

Code 
rmhi <- read_csv("datasets/MEHOINUSA672N.csv", col_names = TRUE) |> 
  rename(date = DATE, median_income = MEHOINUSA672N)
rmhi |> head()

Scales (2/4)

Plot the data

Code 
rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue")

Scale (3/4)

Add labels

Code 
rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue") +
  labs(
    title = "Real Median Household Income in the United States",
    subtitle = "Units: CPI-U-RS Adjusted Dollars, Not Seasonally Adjusted",
    x = "Year",
    y = "2022 CPI-U-RS Adjusted Dollars",
    caption = "Source: U.S. Census Bureau"
  ) +
  theme_linedraw()

What else?

  • Update the y-axis to include commas (or dollar symbols).
  • Set the x-axis and y-axis limits to provide starting and ending points.
  • Customize x-axis and y-axis break points (tick marks).
  • Customize various fonts as needed.

Scale (4/4)

Update the graphic to include the suggestions above

Code 
date_limits = c(as_date("1980-01-01"), as_date("2025-01-01"))
date_breaks = seq(as_date("1980-01-01"), as_date("2025-01-01"), by = "5 years")

rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue") +
  labs(
    title = "Real Median Household Income in the United States",
    subtitle = "Units: CPI-U-RS Adjusted Dollars, Not Seasonally Adjusted",
    x = "Year",
    y = "2022 CPI-U-RS Adjusted Dollars",
    caption = "Source: U.S. Census Bureau"
  ) +
  scale_y_continuous(labels = scales::dollar) +
  scale_x_date(limits  = date_limits, breaks = date_breaks, date_labels = "%Y") +
  theme_linedraw()

Themes (1/2)

Themes give you considerable control over the appearance of your plot.

Code 
date_limits = c(as_date("1980-01-01"), as_date("2025-01-01"))
date_breaks = seq(as_date("1980-01-01"), as_date("2025-01-01"), by = "5 years")

rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue") +
  labs(
    title = "Real Median Household Income in the United States",
    subtitle = "Units: CPI-U-RS Adjusted Dollars, Not Seasonally Adjusted",
    x = "Year",
    y = "2022 CPI-U-RS Adjusted Dollars",
    caption = "Source: U.S. Census Bureau"
  ) +
  scale_y_continuous(labels = scales::dollar) +
  scale_x_date(limits  = date_limits, breaks = date_breaks, date_labels = "%Y") +
  theme_linedraw()

Proposed changes:

  • Increase the font size of the title and make the subtitle italic.
  • Change the font size of the x and y axis text.
  • Move the title, subtitle, and caption to the left.
  • Remove the x-axis and y-axis labels (use the subtitle for y-label).

Look at ?theme

Themes (2/2)

Update the graphic to include the suggestions above

Original visualization:

Code 
date_limits = c(as_date("1980-01-01"), as_date("2025-01-01"))
date_breaks = seq(as_date("1980-01-01"), as_date("2025-01-01"), by = "5 years")

rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue") +
  labs(
    title = "Real Median Household Income in the United States",
    subtitle = "Units: CPI-U-RS Adjusted Dollars, Not Seasonally Adjusted",
    x = "Year",
    y = "2022 CPI-U-RS Adjusted Dollars",
    caption = "Source: U.S. Census Bureau"
  ) +
  scale_y_continuous(labels = scales::dollar) +
  scale_x_date(limits  = date_limits, breaks = date_breaks, date_labels = "%Y") +
  theme_linedraw()

Updated visualization:

Code 
date_limits = c(as_date("1980-01-01"), as_date("2025-01-01"))
date_breaks = seq(as_date("1980-01-01"), as_date("2025-01-01"), by = "5 years")

rmhi |> 
  ggplot() +
  geom_line(aes(x = date, y = median_income), color = "blue") +
  labs(
    title = "Real Median Household Income in the United States",
    subtitle = "CPI-U-RS Adjusted Dollars (2022), Not Seasonally Adjusted",
    x = NULL,
    y = NULL,
    caption = "Source: U.S. Census Bureau"
  ) +
  scale_y_continuous(labels = scales::dollar) +
  scale_x_date(limits  = date_limits, breaks = date_breaks, date_labels = "%Y") +
  theme_linedraw() +
  theme(
    plot.title = element_text(size = 20, face = "bold", hjust = 0),
    plot.title.position = "plot",
    plot.subtitle = element_text(size = 14, face = "italic", hjust = 0),
    axis.title.x = element_blank(),
    axis.text = element_text(size = 12),
    axis.title.y = element_blank(),
    plot.caption = element_text(size = 10, hjust = 0, vjust = 0)
  )

Layout (1/2)

Patchwork!

patchwork

The goal of patchwork is to make it ridiculously simple to combine separate ggplots into the same graphic. As such it tries to solve the same problem as gridExtra::grid.arrange() and cowplot::plot_grid but using an API that incites exploration and iteration, and scales to arbitrarily complex layouts.

Code 
library(patchwork)

Layout (2/2)

Combine the two plots

Code 
p1 <- penguins |> 
  ggplot() +
  geom_boxplot(aes(x = species, y = bill_depth_mm, fill = species), show.legend = FALSE) +
  labs(tag = "A") + theme(plot.tag.location = "panel", plot.tag.position = "topright")

p2 <- penguins |> 
  ggplot() +
  geom_point(aes(x = bill_length_mm, y = bill_depth_mm, color = species)) +
  labs(tag = "B") +theme(plot.tag.location = "panel", plot.tag.position = "topright")

(p1 + p2) + plot_layout(guides = "collect")

Assignment

When Online Content Disappears

Recreate the plot from the article:

Pew Research

Code 
# Font looks close to google font "work sans"  

pew_data <- c(38, 35, 31, 30, 26, 31, 32, 27, 22, 15, 8)
years <- 2013:2023
color = "#ed9d09"

pew_df <- tibble(year = years, percent = pew_data)
pew_df

Sandbox

Code 
library(ggtext)

caption_text = paste(str_wrap("Source: Pew Research Center analysi of a random selection of URLs collected by the Common Crawl web repository (n=999,989) and checked using page and DNS response codes. Web pages defined as inaccessible if they returned a status code of 204, 400, 404, 410, 500, 501, 502, 503, 523 or did not return a valid status code.",
                 width = 86),
        "\n\"When Online Content Disappears\"")

pew_df |> 
  ggplot() +
  geom_line(aes(x = year, y = percent), color = color, linewidth = 3) +
  geom_point(aes(x = year, y = percent), shape = 21 , color = color, fill = "white", size = 4, stroke = 1) +
  geom_text(aes(x = year, y = percent, label = paste0(percent, "%")), nudge_y = 2.2, nudge_x = 0.0, size = 4, color = color, data = filter(pew_df, year == 2013)) +
  geom_text(aes(x = year, y = percent, label = percent), nudge_y = 2.2, nudge_x = 0, size = 4, color = color, data = filter(pew_df, year >= 2014)) +
#  annotate("text", x = 2014.05, y = 5, label = "PEW RESEARCH CENTER", color = "black", size = 4, fontface = "bold") +
  labs(
    title = "38% of webpages from 2013 are no longer accessible",
    subtitle = "% of links from each year that are no longer accessible as of October 2023",
    x = "",
    y = NULL,
    caption = caption_text,
    tag = "PEW RESEARCH CENTER"
    ) +
  theme_minimal() +
  scale_y_continuous(limits = c(5, 42)) +
  scale_x_continuous(limits = c(2012.75, 2023.25), breaks = 2013:2023) +
  theme(
    plot.title = element_markdown(size = 20, face = "bold"),
    plot.title.position = "plot",
    plot.subtitle = element_markdown(size = 14, face = "italic", color = "#666666"),
    plot.caption = element_text(size = 12, face = "italic", hjust = 0, color = "#666666"),
    plot.caption.position = "plot",
    axis.text.x = element_text(size = 12),
    axis.line.x = element_line(color = "black"),
    axis.ticks.x = element_line(color = "black"),
    axis.text.y = element_blank(),
    plot.tag.location = "plot",
    plot.tag.position = "bottomright",
    plot.tag = element_text(size = 12, face = "bold", color = "black"),
    panel.grid = element_blank()
  )

End of Module 4

References

Applied Statistical Techniques